Can the Application of Environmentally Friendly Fertilisers Reduce Agricultural Labour Input? Empirical Evidence from Peanut Farmers in China

Abstract

Environmentally friendly fertilisers (EFFs) can improve the quality of cultivated land, purify the soil environment, and promote reduction in the amounts of fertiliser applied by improving efficiency. However, few studies have analysed the spillover effects of EFF applications on agricultural labour inputs. Hence, this study discusses the impact of EFFs on agricultural labour input, using the propensity score matching method based on the micro-survey data of peanut growers in the main producing areas in China. The results showed that EFFs have a labour-saving advantage, with a significant average reduction in the number of labourers’ input and labour days in agriculture production of 0.127 persons/mu and 0.601 days/mu at the 1% to 10% significance level. Additionally, EFFs improve yield and revenue but significantly raise production costs, ultimately increasing net revenue for farmers. The mechanism of the labour-saving effect is the capability of EFFs to reduce the amount and frequency of fertilisers applications, the amounts of pesticides applied, and irrigation water consumption by their technical characteristics and farmland’s ecological environment enhancement. Accordingly, the time effect strengthens the ecological regulation function and the application reduction effect of EFFs, further decreasing agricultural labour inputs. At the same time, the application of EFFs contributes to the adoption of mechanical deep tillage and fertilisation technology (MDTFT), thereby reducing fertiliser application and ultimately improving the labour-saving effect of EFFs. Heterogeneity analysis revealed that the labour-saving effect of EFFs is more obvious for farmers operating with a larger planting scale. To improve the labour-saving effect of EFFs, the findings imply that the application years of EFFs should be appropriately extended and the MDTFT should be promoted according to local conditions.

1. Introduction

Chemical fertilisers have made outstanding contributions towards improving land-use efficiency and ensuring the supply of agricultural products [1]. However, the application of chemical fertilisers in China has exceeded the optimal range of the coordinated development of agricultural production and the ecological environment [2]. This situation not only leads to agricultural non-point source pollution, soil compaction, and acidification but also increases fertiliser and labour costs because of excessive fertilisation, which weakens the price competitiveness of agricultural products in China [3,4]. In 2021, the ‘14th Five-Year Plan for Green Development of National Agriculture’ clearly pointed out that the country should drive the growth of the agricultural economy through green inputs, further improve cultivated land quality (CLQ) and fertiliser utilisation efficiency, and strengthen the management of fertiliser non-point source pollution [5]. In 2022, the 20th National Congress Report proposed strengthening the construction of high-standard farmland. As a multi-functional fertiliser product, environmentally friendly fertilisers (EFFs) have become essential means to promote the green development of agriculture [6].

EFFs are novel fertiliser products made with novel materials, such as biochar, and through processes, including blending, adsorption, coating, and blending and granulation methods. EFFs have characteristics, including compounding, high fertiliser efficiency, long-lasting fertility, and low environmental risk. They are synthesized with organic matter, inorganic nutrients, and beneficial microorganisms using biochar as the substrate, including biochar-based organic fertilisers (BOFs), biochar-based formula fertilisers (BFFs), biochar-based slow-release fertilisers (BSRFs), and biochar-based bacterial fertilisers (BBFs). Compared with traditional chemical fertilisers (TCFs), EFFs are multi-functional fertilisers that can increase crop yield and have strong positive externalities [1]. Specifically, EFFs locked in soil nutrients and water established long-lasting fertilisation effects and thus reduced the amount and frequency of topdressing [7,8,9]. Simultaneously, EFFs loosened compacted soil, improved the CLQ, purified the soil microenvironment, and reduced the accumulation of harmful microorganisms [10,11], thereby reducing the application of chemical fertilisers and pesticides as the frequency of tillage. However, realising the above functions requires a specific period. The continuous long-term application of EFFs can fully regulate soil physicochemical properties, thus strengthening the ecological regulation function [12,13]. Hence, EFFs are typical fertiliser products, which exert an inter-period application effect. Under the above mechanism, the impact of EFFs on the CLQ, fertiliser, and pesticide application and irrigation water consumption (IWC) will change the labour inputs in cultivation, fertilisation, irrigation, and pest control.

Previous studies have discussed the green technology adoption behaviour of farmers from the perspectives of credit service, technical training, technical extension services, and social capital [14,15,16,17] as well as the influence of land transfer, operation scale, rural e-commerce, and social networks on the intensity of fertiliser application [2,18,19,20]. Other studies further revealed the effect of agricultural technology adoption on agricultural labour inputs [21,22,23,24,25]. However, research on the effects of EFFs on agricultural labour inputs is still absent. The spillover effect of technical characteristics and the ecological regulating function of the application of EFF on agricultural labour input has not been determined. The influence of the inter-period application effect of EFFs on labour inputs also lacks systematic analyses. In addition, EFFs can affect the adoption of mechanical deep tillage and fertilisation technology (MDTFT), which can further influence agricultural labour inputs. Nevertheless, the discussion in this section is still insufficient in existing studies.

Peanut farmers were investigated in this study. Peanuts are labour-intensive crops. The labour cost of peanuts in 2020 was 685.03 CNY/mu, accounting for more than half of the production cost, whereas the average labour cost of rice, corn, and wheat was 398.74 CNY/mu, only 58.2% of the labour cost of peanuts. In contrast, the average labour costs for rice, maize, and wheat accounted for 36.87% of the total costs. The return on peanuts was 1.63 times higher than the average return on rice, maize, and wheat. Peanuts, as a high-return crop, require a greater labour investment than other food crops. In recent years, the efficiency of fertiliser applied to peanuts has continued to decrease due to the degradation of CLQ caused by excessive fertiliser application. Since 2016, the peanut yield per unit of fertiliser applied has decreased by an average of 3% per year, severely limiting the peanut yield growth. The above data were obtained from the National Agricultural Product Cost Information Compilation. Furthermore, farmers need to increase the application of fertilisers to maintain stable yields, further raising the labour demand for peanut production. Therefore, peanut, as a labour-intensive fertiliser product, has been more severely affected than other crops.

To fill a gap in the existing literature, this paper discusses the impact of EFFs on peanut production labour input and the mechanism according to the inter-period application effect and multi-functional feature of EFFs. EFFs are a multi-functional fertiliser product; therefore, their technical characteristics and farmland ecological regulation function collectively influence agricultural labour input. EFFs also have the property of inter-period application effects. The regulating effect on farmland ecosystems will be enhanced with longer application years. Thereby, the time effect will deepen their influence on agricultural labour input. In addition, this study considers the impact of EFFs on adopting MDTFT and thus discusses how MDTFT affects labour inputs. Exploring the above issues will deepen the understanding of the environmental benefits and labour substitution effects of EFFs and, simultaneously, provide a realistic path to reduce labour costs and alleviate the pressure of labour demand for peanut production.

The remainder of this paper is organised as follows:

(1) Based on the multi-functionality of EFFs, this study theoretically determined the application reduction effects, fertilisation technology effects, and time effects of EFFs and further analysed the influence of the above effects on agricultural labour inputs.

(2) The propensity score matching (PSM) method and endogenous switching regression (ESR) model were used to eliminate the selection bias of observable and unobservable factors and identify the net effect of EFFs on agricultural labour inputs. Furthermore, the mechanism was tested empirically to provide data supporting the theoretical derivation.

(3) Scale-heterogeneity effects of EFFs on agricultural labour inputs are considered.

2. Analytical Framework

This study constructed an analytical framework for the impact of EFFs on labour input based on the multi-functionality and inter-period application effect of EFFs. The specific impact paths are shown in Figure 1.

2.1. Application Reduction Effect

Agroecosystems are complex networks consisting of biological communities and other abiotic factors. The adoption of new agricultural technologies not only affect a single production factor but also has spillover effects on other factors through agroecosystems [26,27]. As multi-functional fertilisers, EFFs not only supply nutrients, such as nitrogen, phosphorus, and potassium, for crop growth but also have several ecological functions, such as alleviating soil compaction and acidification caused by over-fertilisation, improving soil fertility, and optimising the soil microenvironment. Therefore, the technical characteristics and improvement effect on the farmland ecological environment of EFFs reduce the fertiliser application amount (FAA) and have a systematic impact on the input of other production factors, such as pesticides and irrigation water. Such results eventually change the labour input in fertilisation, disease and pest control, and irrigation.

(1) There are several effects of reducing the application of fertiliser. Firstly, EFFs match the nutrient release rate with the nutrient demand of the crop and prolong the fertiliser effect time, thus reducing the amount and frequency of topdressing [8]. EFFs also lock soil nutrients and regulate soil aggregate structures through biochar to improve the CLQ, thus reducing the FAA [10]. Finally, EFFs significantly reduce the labour input in fertilisation.

(2) There are also notable water-saving effects in reducing the application of fertiliser. Excessive fertilisation leads to soil compaction, which make it difficult for water to penetrate the soil. Compared with TCFs, EFFs loosen the consolidated soil and adjust the soil porosity and have a strong adsorption effect to lock soil moisture, thus improving the physicochemical properties and moisture retention capacity of the soil as well as reducing the IWC [9]. Therefore, EFFs reduce labour input in irrigation and cultivation.

(3) Reducing the application of pesticide results in some additional effects. Excessive fertilisation deteriorates the soil environment and increases the accumulation of harmful bacteria. EFFs purify the soil microenvironment, increase the number of beneficial microorganisms, and reduce the reproduction ability of harmful bacteria and the incidence of diseases and pests [28,29]. Meanwhile, EFFs further reduced pesticide application by improving plant disease resistance [30]. Ultimately, EFFs reduce the labour input in pest and disease control.

2.2. Fertilisation Technology Effect

As a multi-functional fertiliser product, EFFs reduce the FAA, purify the soil microenvironment, and improve soil water and fertiliser retention capacity. However, different fertilisation methods will affect the performance of these functions: the traditional manual fertilisation method is relatively simple, and farmers mainly apply fertilisers by spreading a one-time application. Fertilisers are applied on the soil surface with severe nutrient loss, and artificial topdressing is required at a later stage, leading to high labour demand [31]. In comparison, MDTFT not only directly replaces artificial fertilisation but also applies fertiliser to the easily absorbed parts of crops below the surface, which reduces nutrient loss and improves the utilisation efficiency of fertiliser [32]. Consequently, MDTFT improves the effect of fertiliser reduction and reduces the amount and frequency of topdressing, thus saving agricultural labour input. Farmers who adopt EFFs often experience the degradation of CLQ and a reduction in reduced agricultural output per fertiliser unit. Compared with farmers who apply TCFs, these farmers are more motivated to improve fertiliser utilisation efficiency and crop yield by improving soil fertility and therefore prefer to adopt MDTFT for fertilisation.

2.3. Time Effect

The reduction in the application of fertilisers, pesticides, and other production factors depends on the progress of agricultural technology, and the function of new technologies is constrained by the years of application [33]. In terms of product attributes, the application effect of EFFs has inter-period characteristics, and the improvement effect on the CLQ and soil microenvironment may not be obvious in the short term. The soil regulation function of EFFs needs a certain time cycle to play out, and with the extension of application years, the fertiliser and pesticide application reduction effect and water-saving effect of EFFs will be improved [12,13]. Therefore, when analysed from the time dimension, the impact of EFFs on agricultural labour inputs in the initial application process is limited, and the labour-saving effects of EFFs will gradually manifest in the long term.

In general, the application reduction effect, time effect, and effect of fertilisation technology of EFFs jointly determine the net effect on agricultural labour inputs, whereas the time effect and effect of fertilisation technology will deepen the application reduction effect of EFFs and further improve their labour-saving effect. To test the validity of the previous logical inference, an analytical framework for the effect of EFFs on labour substitution was constructed in this study (Figure 1) and verified via empirical analysis.

3. Materials and Methods

3.1. Data Sources

Relying on the Industrial Economy Research Office of the National Modern Agriculture (Peanut) Industry Technology System, this study conducted field research on peanut farmers in the main peanut-producing areas of China. Based on the consideration of economic development level, regional distribution, and the peanut-planting situation, this study selected the main peanut-producing areas, including the Huang-Huai-Hai region, northeast region, Yangtze River Basin, and southern region. These areas can better represent the national peanut production situation. A stratified random sampling method was used for field research, relying on experimental stations in each key peanut-producing area. In total, 19 experimental stations were selected according to the application of EFFs, and 5 sample counties were stratified for each experimental station; 4–8 peanut farmers were randomly selected in each sampled county, and 546 farmer questionnaires in total were collected. After post-summary and data cleaning, 528 valid questionnaires were obtained; the questionnaire efficiency reached 96.7%. The field research on peanut farmers was conducted with the participation of graduate students and by the means of household research, which ensured the validity of the data sources of this study. This study focused on the peanut production by peanut farmers, including the characteristics of farmers and households, peanut production and operation, the farming environment, and the application behaviour of EFFs. The homogeneity of the farmers is that the study was conducted with farmers in the EFF promotion trial area. The EFFs were sold at agricultural production service stations in the pilot area and were one of the fertiliser options most available to farmers compared with other non-promoted areas. Moreover, farmers in the pilot area could learn the application methods and characteristics of EFFs sufficiently to have the objective conditions to adopt them. Additionally, farmers chosen in this study have been growing peanuts for years, possess significant experience in peanut cultivation and management techniques, and have a realistic basis for EFF promotion. The heterogeneity of the production and management characteristics of the sample farmers in the treated and untreated groups is shown in

Table 1. Descriptive statistics of variables.

Variable	Unit	EFFs		TCFs		Difference	Standard Error (SE)
		Mean	SE	Mean	SE
Independent Variables
Individual and household characteristics
Gender	1 = male, 0 = female	0.923	(0.267)	0.938	(0.242)	−0.014	(0.022)
Age	year	50.230	(8.782)	54.099	(9.082)	−3.870 ***	(0.796)
Education level	year	9.938	(2.249)	9.345	(2.247)	0.593 ***	(0.200)
Whether there are village cadres in the household	1 = Yes; 0 = No	0.220	(0.415)	0.211	(0.409)	0.009	(0.037)
Whether there are party members in the household	1 = Yes; 0 = No	0.351	(0.478)	0.377	(0.485)	−0.026	(0.043)
Whether it is a minority household	1 = Yes; 0 = No	0.120	(0.325)	0.087	(0.282)	0.033	(0.027)
Proportion of children under 6 years old	—	0.046	(0.097)	0.083	(0.215)	−0.036 **	(0.016)
Proportion of people over 65 years old	—	0.065	(0.140)	0.129	(0.253)	−0.065 ***	(0.019)
Per capita income of household labourer	CNY	5.310	(0.141)	3.495	(0.110)	1.814 ***	(0.177)
Operations characteristics
Years of peanut cultivation	years	18.000	(12.514)	21.975	(22.511)	−3.975 **	(1.706)
Peanut sown area	mu	66.421	(149.791)	30.093	(102.642)	36.328 ***	(10.957)
Type of operation organisation	1 = farmer with organised operations, 0 = ordinary farmers	0.440	(0.498)	0.205	(0.404)	0.235 ***	(0.039)
Slope of land	1 = flat land, 0 = non-flat land	0.785	(0.412)	0.677	(0.468)	0.108 ***	(0.040)
Degree of mechanisation	Number of mechanised links	4.622	(2.284)	3.550	(2.337)	1.072 ***	(0.206)
Instrumental variable
Adoption rate of EFFs at village level	—	0.221	(0.095)	0.164	(0.091)	0.057 ***	(0.008)
Dependent variable
Number of labourers per mu	person/mu	0.371	(0.434)	0.522	(0.548)	−0.151 ***	(0.045)
Labour days per mu	day/mu	5.640	(1.911)	6.166	(3.647)	−0.526 *	(0.274)
FAA (converted)	kg/mu	16.516	(4.036)	23.427	(5.134)	−6.911 ***	(0.835)
FAF	frequency	1.306	(0.511)	2.466	(0.622)	−1.160 ***	(0.052)
PAA	g/mu	157.262	(194.678)	201.680	(159.074)	−44.419 **	(20.022)
IWC	m3/mu	56.946	(60.414)	69.218	(101.173)	−12.272	(7.767)
MDTFT adoption decision	1 = Yes; 0 = No	0.335	(0.125)	0.093	(0.054)	0.242 **	(0.121)

Note: Robust standard errors are in parentheses. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively, and the same below.

.

3.2. Model Setting

Influenced by farmers’ individual and household characteristics and the environmental factors of cultivated land, the EFF adoption decisions of farmers were not completely exogenous and sample selection biases appeared, which inevitably affected the accuracy of the estimation results. In this study, the PSM method and the ESR model were used to remove bias in the estimation results due to observable and unobservable variables, respectively. Thus, the treatment effects of EFFs on agricultural labour inputs were evaluated, scientifically validating the results of this study.

(1) PSM method

The PSM method eliminated bias in the estimation results due to observable variables. Thus, the treatment effect of EFFs on agricultural labour inputs can be evaluated.

Firstly, the selection equation of farmers’ EFFs is as follows:

$$D_i={\beta }_0+{\beta }_1{X}_i+{\epsilon }_i$$

(1)

In Equation (1), D_i is the binary choice variable for EFF adoption decisions; X_i is the exogenous explanatory variable affecting farmers’ decisions, including their individual, household and operation characteristics; and ε_i is the random disturbance term. This study used the logit model to calculate the propensity scores of farmers and match the sample of farmers in the adoption and non-adoption groups. For the robustness of estimation results, four mainstream matching methods were used to match the samples in this study:

① nearest-neighbour matching (NNM), a one-to-one matching method used to obtain the nearest sample of propensity scores;

② calliper matching (CM), where the default 0.05 radius is used to match the samples;

③ kernel matching (KM), the default kernel function was selected to obtain the nearest matched sample;

And ④ the bootstrap method (BM), a one-to-one matching method used to repeat sampling 500 times to obtain the mean value. Subsequently, the selected covariates, X_i, were tested for equilibrium to ensure that there was no significant difference between the characteristics of farmers in the treatment and control groups.

Secondly, the average treatment effect on treated (ATT) of EFFs on labour inputs for peanut production was estimated and could be expressed as follows:

$$ATT=\frac{1}{N}{\displaystyle {\sum }_{i:D_i^j=1}(y_i^1-{\widehat{y}}_i^0)}$$

(2)

In Equation (2), N is the number of samples in the treatment group, $y_i^1$ is the labour input in the treatment group, and ${\widehat{y}}_i^0$ is the labour input in peanut production of farmers in the control group matched with the treatment group.

(2) ESR model

The PSM method can only deal with sample self-selection bias caused by observable factors and cannot avoid the possible influence caused by unobserved factors. For this reason, this study used the ESR model to address bias in the estimation results due to unobserved factors by introducing an inverse Mills ratio. Simultaneously, an instrumental variable (IV) was used to address the endogeneity of farmers’ EFF adoption decisions. The advantages of ESR model were as follows: On the one hand, it can solve the sample selection bias and endogeneity problems caused by observable and unobserved factors; on the other hand, the counterfactual condition can be constructed to divide the sample farmers into treatment and control groups, and the average treatment effect (ATE) of EFFs on agricultural labour inputs can be obtained for the treated and untreated groups of farmers, respectively. Thus, Equation (1) becomes:

$$D_i={\beta }_0+{\beta }_1{X}_i+{\beta }_2{I}_i+{\epsilon }_i$$

(3)

In Equation (3), I_i is the IV. To solve the bias caused by omitted variables and unobserved factors, the inverse Mills ratio (λ) calculated by using Equation (3) was introduced into the result equation, and the resultant equation is as follows:

Farmers applying EFFs (treatment group):

$$L_T={\alpha }_0+{\alpha }_1{X}_i+{\sigma }_{T_v}{\lambda }_T+{\epsilon }_T$$

(4)

Farmers not applying EFFs (control group):

$$L_U={\alpha }_0+{\alpha }_1{X}_i+{\sigma }_{U_v}{\lambda }_U+{\epsilon }_U$$

(5)

In Equations (4) and (5), L is the number of labourers or labour days for peanut production per mu; λ_T and λ_U represent the influence of unobserved factors (e.g., the ability and intelligence level of farmers) on the treatment group and control group, respectively; and σ_Tυ and σ_Uυ represent the covariance of random disturbance terms of the treatment group and control group, respectively.

Due to the differences in individual, household, and business characteristics between the adoption and non-adoption groups, a direct comparison of the two groups would lead to biased estimation results. Based on this, counterfactual groups were constructed for the adoption and non-adoption groups, respectively, for the adoption group if they did not adopt EFFs, and for the non-adoption group if they did adopt EFFs. Based on the counterfactual framework, the ATE of EFFs on agricultural labour input was estimated under the same matching conditions using the real situation of farmers and the counterfactual situation.

Farmers adopting EFFs (treatment group):

$$E[L_{iT}|D_i=1]={\alpha }_T{X}_{iT}+{\sigma }_{T_v}{\lambda }_T$$

(6)

Farmers not adopting EFFs (control group):

$$E[L_{iU}|D_i=0]={\alpha }_U{X}_{iU}+{\sigma }_{U_v}{\lambda }_U$$

(7)

Farmers adopting EFFs if they did not adopt (counterfactual to the treatment group):

$$E[L_{iU}|D_i=1]={\alpha }_U{X}_{iT}+{\sigma }_{U_v}{\lambda }_T$$

(8)

Farmers not adopting EFFs if they did adopt (counterfactual to the control group):

$$E[L_{iT}|D_i=0]={\alpha }_T{X}_{iU}+{\sigma }_{T_v}{\lambda }_U$$

(9)

In conclusion, the ATT of the actual adoption of EFFs on the agricultural labour input of farmers is as follows:

$$ATT=E[L_{iT}|D_i=1]-E[L_{iU}|D_i=1]=({\alpha }_T-{\alpha }_U)X_{iT}+({\sigma }_{T_v}-{\sigma }_{U_v}){\lambda }_T$$

(10)

Similarly, if farmers in the actual non-adoption group adopted, then the average treatment effect on untreated (ATU) of EFFs on the agricultural labour input will be as follows:

$$ATU=E[L_{iT}|D_i=0]-E[L_{iU}|D_i=0]=({\alpha }_T-{\alpha }_U)X_{iU}+({\sigma }_{T_v}-{\sigma }_{U_v}){\lambda }_U$$

(11)

3.3. Variable Selection

(1) Outcome variables: Number of labourers per mu and labour days per mu.

(2) Mechanism variables: FAA, fertiliser application frequency (FAF), pesticide application amount (PAA), IWC, MDTFT adoption decision, and application years of EFFs were selected as mechanism variables. Among them, IWC was calculated by the average water cost per mu and the local water price.

(3) Treatment variable: The adoption decision of farmers for EFFs. The ‘fertiliser application category’ of the questionnaire identifies this variable. The value of treatment variable was ‘1’ if the fertiliser application category included one or more of the following fertilisers: BOFs, BFFs, BSRFs, and BBFs. Otherwise, a value of ‘0’ was assigned.

(4) Control variables: Considering the impact of farmers’ endowments and the heterogeneity of the production and operation environment, this study incorporated individual, household, and operation characteristic variables of farmers into the model, as shown in Table 1.

(5) Instrumental variable: Due to the peer effects, farmers’ behaviour is influenced by their group. Therefore, this study selected the adoption rate of EFFs of village-level farmers as the IV following the research of Bentolila et al. [34]. This IV is used to measure the impact of the adoption decision of farmers in the same village on individual decision-making but does not affect the agricultural labour input. The descriptive analysis of variables is shown in Table 1.

4. Results

4.1. Baseline Results

This section describes the common domain and balance test for PSM. Then, the ATT of EFFs on labour input in peanut production was analysed.

4.1.1. Common Domain and Balance Test

According to the matching results, 501 farmers in the 528 samples were in the common domain, and only 27 were outside the common domain. The amount of sample loss in the matching process was small, and the overall impact on the sample after elimination was insignificant, which ensured the adequacy of the sample size in the common domain. Among them, the sample sizes of farmers in the treatment and control groups were 95 and 406, respectively, which could better meet the need for sample size in discussing the ATT of EFFs on labour input.

Table 2. Balance test before and after matching.

Method	R2	p-Value	Mean Deviation	Median Deviation (%)	Overall Deviation (%)
No match	0.237	0.000	27.6	27.5	64
NNM	0.040	0.122	10	8	28
CM	0.046	0.126	9.3	6.4	36
KM	0.046	0.056	9.5	6.4	36
BM	0.039	0.262	9.7	7.1	36

reports the balance test results using four primary matching methods: NNM, CM, KM, and BM. After matching, R² decreased from 0.237 to 0.039–0.046, indicating an improved group balance between the treatment and control groups; the p-value improved from 0.00 to 0.056–0.262; the mean deviation decreased from 27.6 to 9.3–10; the median deviation decreased from 27.5 to 6.4–8; and the overall deviation decreased from 64 to 28–36, indicating a significant reduction in the overall deviation of the matched samples.

In this paper, the deviations of the independent variables for the treated and untreated groups after matching using NNM, CM, KM, and BM were examined, respectively. As the results of the above matching method tests were generally consistent, this paper shows the differences in the independent variables after matching using NNM as an example (

Table 3. Deviation and significance of independent variables in the treated and untreated groups after matching (NNM).

	Variables	Deviation of the Treatment Group from the Control Group
		Deviation (%)	T-Value	p-Value
Individual and household characteristics	Gender	4.4	0.42	0.673
	Age	−4.5	−0.43	0.665
	Education level	8.7	0.90	0.370
	Whether there are village cadres in the family	6.9	0.65	0.517
	Whether there are party members in the family	−6.6	−0.62	0.533
	Whether it is a minority household	10.8	1.23	0.218
	Proportion of children under 6 years old	−2.1	−0.31	0.755
	Proportion of people over 65 years old	−12.4	−1.34	0.182
	Per capita income of household labourer	2.6	0.81	0.420
Operations characteristics	Years of peanut cultivation	12.5	1.11	0.267
	Peanut sown area	5.7	0.65	0.515
	Type of operation organisation	12.5	1.11	0.267
	Slope of land	8.5	0.84	0.404
	Degree of mechanisation	16.8	1.51	0.132
	Adoption rate of EFFs	8.7	0.82	0.412

). The results show that each explanatory variable after matching is not significantly different between the experimental and control groups. Overall, using the PSM method eliminated significant differences between the experimental and control groups and reduced sample selection bias.

4.1.2. Results of Average Treatment Effect

The ATT of EFFs on agricultural labour input was measured using the PSM method; the estimated results are shown in

Table 4. ATT for the PSM method.

	Matching Method	ATT	SE	T-Value
Number of labourers per mu	NNM	−0.147 ***	0.041	−3.56
	CM	−0.151 ***	0.045	−3.36
	KM	−0.105 *	0.062	−1.69
	BM	−0.102 ***	0.030	−3.41
	Mean	−0.127	—	—
Labour days per mu	NNM	−0.529 *	0.274	−1.93
	CM	−0.530 **	0.270	−1.96
	KM	−0.666 **	0.331	−2.01
	BM	−0.679 ***	0.204	−3.32
	Mean	−0.601	—	—

Note: ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

. This study found that EFFs significantly reduced the number of labourers and labour days at a significance level of 1% to 10% using various matching methods. The estimation results for the different matching methods are consistent, proving that the measurement results of this study are robust and reliable. Based on this result, the mean values of the four matching methods were used to represent the ATT of EFFs on labour inputs.

As shown in Table 4, the average ATT of EFFs on the number of labourers per mu in peanut production was −0.127, indicating that EFFs could reduce the number of labourers per mu in peanut production by 0.127, a reduction of 24.3%. The average ATT of EFFs on labour days per mu in peanut production was −0.601, indicating that the application of EFFs could save labour days in peanut production by 0.601 days/mu on average, a decrease of 9.7%. On the whole, compared with TCFs, EFFs not only can promote the reduced and efficient use of fertilisers but also have apparent labour-saving advantages at a significance level of 1% to 10%, which can effectively alleviate the pressure of labour costs due to the increase in labour price.

4.2. Robustness Checks

In the ESR model, the endogeneity of farmers’ EFF adoption decisions needs to be examined because such adoption decisions are influenced by the perceptions, abilities, and environment of the farmers. The Durbin–Wu–Hausman (DWH) statistic and p-value show that the hypothesis of exogenous variables was rejected at the 1% and 10% confidence levels, indicating the existence of endogeneity in adopting EFFs (

Table 5. Endogeneity test.

		Number of Labourers per mu	Labour Days per mu
DWH	Statistic	12.8232	3.739
	p-value	0.0003	0.053
Weak instrumental variable test	Cragg–Donald Wald F Statistic	29.349	29.346
	Stock–Yogo critical value	16.380	16.380

). On this basis, the weak IV test was conducted on the selected IV. The Cragg–Donald Wald F statistics were 29.349 and 29.346, respectively, greater than the Stock–Yogo critical value of 16.380. Therefore, the IVs selected in this study did not have weak IV problems, and the selected IVs were reasonably valid.

Table 6. ATE of the ESR model.

	Group	Decision		ATE	SE
		Adopted	Not Adopted
Number of labourers per mu	Adopted group	0.341	0.433	−0.092 ***	0.029
	Not adopted group	0.480	0.512	−0.032 ***	0.007
Labour days per mu	Adopted group	5.551	6.066	−0.515 ***	0.121
	Not adopted group	5.858	6.165	−0.307 ***	0.092

Note: *** denote significance at the 1% level, respectively.

reports the estimation results of the ESR model. The study found that the values of the ATE of EFFs on the number of labourers in the adopted and unadopted groups were −0.092 and −0.032, respectively, indicating that the number of labourers per mu in the adopted group was reduced by 0.092 compared with the counterfactual case. The number of labourers per mu in the unadopted group could be saved by 0.032 if farmers adopted EFFs. The values ATE of EFFs on the labour days of farmers in the adopted and unadopted groups were −0.515 and −0.307, respectively, indicating that the labour days per mu in the adopted group were reduced by 0.515 days compared with the unadopted group. The labour days per mu in the unadopted group were reduced by 0.307 days if EFFs were adopted. A comparison between the ATE of the PSM model and ESR model indicated that the two estimation results were consistent in the coefficient magnitude and direction and that the previous findings still held after changing the estimation method. Thus, the robustness of the baseline estimation results was proven.

4.3. Mechanism Tests

The application reduction effect includes reducing the FAA, FAF, PAA, and IWC. The time effect deepens the application reduction effect. The fertilisation technology effect is the impact of farmers’ EFF adoption on MDTFT adoption. Consequently, agricultural labour inputs are saved through the above pathways.

Table 7. Mechanisms of the reduction effect and the fertilisation technology effect on labour input.

	FAA	FAF	PAA	IWC	MDTFT Adoption Decision	FAA
	(1)	(2)	(3)	(4)	(5)	(6)
EFF adoption decision	−5.927 ***	−1.200 ***	−41.742 *	−12.569 ***	0.715 ***
	(0.285)	(0.051)	(21.379)	(2.376)	(0.151)
MDTFT adoption decision						−0.583 ***
						(0.211)
Control variables	Control	Control	Control	Control	Control	Control
Sample size	528	528	528	528	528	528
R2	0.511	0.549	0.144	0.202	—	0.518
Wald test value	—	—	—	—	141.94 ***	—
	Number of labourers per mu
	(7)	(8)	(9)	(10)	(11)	(12)
EFF adoption decision	−0.123 ***	−0.139 ***	−0.105 ***	−0.182 ***	−0.102 ***	−0.083 ***
	(0.046)	(0.053)	(0.011)	(0.068)	(0.010)	(0.012)
FAA	0.008 **					0.004 *
	(0.004)					(0.002)
FAF		0.046 *				0.033 ***
		(0.027)				(0.012)
PAA			0.0003 ***			0.0003 ***
			(0.0001)			(0.0001)
IWC				0.0002		0.0002
				(0.0002)		(0.0003)
MDTFT adoption decision					−0.018 **	−0.025 *
					(0.008)	(0.014)
Control variables	Control	Control	Control	Control	Control	Control
Sample size	528	528	528	528	528	528
Wald test value	367.81 ***	285.87 ***	246.00 ***	373.76 ***	372.71 ***	247.57 ***
	Labour days per mu
	(13)	(14)	(15)	(16)	(17)	(18)
EFF adoption decision	−0.608 ***	−0.626 **	−0.561 *	−0.628 ***	0.579 *	−0.550 *
	(0.071)	(0.287)	(0.321)	(0.244)	(0.307)	(0.285)
FAA	0.062 **					0.084 **
	(0.003)					(0.028)
FAF		0.677 *				0.567 **
		(0.390)				(0.274)
PAA			0.0009 *			0.0008 *
			(0.0005)			(0.0005)
IWC				0.003 ***		0.004 ***
				(0.001)		(0.001)
MDTFT adoption decision					−0.479 ***	−0.422 ***
					(0.175)	(0.159)
Control variables	Control	Control	Control	Control	Control	Control
Sample size	528	528	528	528	528	528
Wald test value	549.05 ***	542.11 ***	466.93 ***	558.05 ***	561.3 ***	500.95 ***

Note: ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

reports the mechanisms of the application reduction and fertilisation technology effects on labour input.

Table 8. Mechanisms of time effects on agricultural labour inputs (NNM).

	Application Year	ATT	SE	T-Value
FAA	2 years or less	−4.926 ***	0.642	−7.67
	More than 2 years	−6.893 ***	1.563	−4.41
PAA	2 years or less	−32.823	21.633	−1.52
	More than 2 years	−52.419 **	23.597	−2.22
IWC	2 years or less	−10.474	25.511	−0.41
	More than 2 years	−16.052 *	8.257	−1.94

Note: ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

reports the deepening effect of the time effect on the application reduction effect of EFFs.

4.3.1. Application Reduction Effect and Fertilisation Technology Effect

The estimation results in columns (1)–(5) in Table 7 show that the adoption of EFFs has a significant negative effect on FAA, FAF, PAA, and IWC, and a significant positive effect on MDTFT adoption. The adoption of EFFs has a significant effect on fertiliser application reduction as well. Meanwhile, as a multi-functional fertiliser product, EFFs also purify the soil microenvironment, reduce the incidence of pests and diseases, and improve the soil’s ability to retain water and fertiliser, thereby further reducing PAA and IWC. In addition, farmers who adopt EFFs tend to have a stronger motivation to improve soil fertility and fertiliser utilisation efficiency, which induces them to adopt MDTFT for fertilisation. As shown in columns (7)–(11) and (13)–(17), FAA, FAF, PAA, and IWC had significant positive effects on the number of labourers and labour days per mu, whereas MDTFT significantly negatively affected the number of labourers and labour days per mu, controlling for the individual, household, and operation characteristics. The estimation results incorporating all mechanism variables in columns (12) and (18) showed that the coefficients of FAA, FAF, PAA, and IWC were significantly positive. In contrast, the coefficient of MDTFT was significantly negative. Meanwhile, the absolute values of the coefficients of EFFs regarding the number of labourers decreased from 0.102–0.182 in columns (7)–(11) to 0.083 in column (12), and the absolute values of the coefficients of EFFs regarding labour days decreased from 0.561–0.628 in columns (13)–(17) to 0.550 in column (18). This result confirmed that EFFs achieved labour-saving effects by FAA, FAF, PAA, and IWC and induced farmers to adopt MDTFT.

This study likewise examined the effect of MDTFT on FAA. The estimation results in column (6) of Table 7 show that the coefficient of MDTFT adoption by farmers on FAA was significantly negative, indicating that adopting MDTFT reduced the FAA. In addition, the study confirmed that adopting EFFs induced farmers to adopt MDTFTs, thus reducing agricultural labour inputs. At the same time, adopting MDTFTs further contributed to fertiliser reduction, which improved the labour-saving effect. Therefore, EFFs achieved a self-deepening of the labour-saving effect.

4.3.2. Time Effect

The estimation results in Table 8 showed that the FAA decreased by 4.926 kg/mu for farmers who had applied EFFs for two years or less, while the PAA and IWC did not change significantly. For farmers who applied EFFs for more than two years, the FAA decreased by 6.893 kg/mu, and the PAA and IWC decreased by 52.419 and 16.052 m³/mu, respectively. This outcome indicated that the fertiliser and pesticide application reduction and water-saving effects of EFFs gradually increased with the extension of application years. Finally, the time effect deepened the application reduction effect and further improved the substitution effect of EFFs on agricultural labour input.

4.4. Scale Heterogeneity

A significant difference existed between smallholder and large-scale peanut production characteristics. Large-scale households, whether from production technology or management experience, were better than smallholder farmers. In this section, based on the two-thirds quantile of peanut cultivation area of farmers in the sample, farmers were split into small- and medium-sized peanut farmers of 20 mu and below and large-scale peanut farmers of 20 mu and above. The NNM method was used to estimate the scale heterogeneity of the impact of EFFs on agricultural labour inputs. The estimated results are shown in

Table 9. Results of heterogeneous cohort differences (NNM).

Classification Criteria	Number of Labourers per mu		Labour Days per mu
	ATT	SE	ATT	SE
Small- and medium-scale farmers	−0.109 *	0.057	−0.485 *	−0.279
Large-scale households	−0.145 **	0.066	−0.761 **	0.310

Note: **, and * denote significance at the 5%, and 10% levels, respectively.

.

The number of labourers and labour days were reduced by 0.145 persons/mu and 0.761 days/mu, respectively, after EFFs were applied by the large-scale households, while the labour-saving effect of the application of EFFs by small- and medium-scale households was lower, with the number of labourers and labour time reduced by 0.109 persons/mu and 0.485 days/mu, respectively. This result indicated that the labour-saving effect of EFFs for large-scale households was significantly higher than that for small- and medium-scale farmers both in terms of labour days and number of labourers. With the expansion of the planting scale, the field management ability and fertiliser application techniques of farmers were improved and the reduction effect of EFFs and the regulation of soil fertility were continuously highlighted. Finally, the advantage of the economy of scale amplified the labour-saving effect of EFFs.

4.5. Further Analysis

This study confirms that EFFs reduce agricultural labour inputs. Therefore, how will the labour-saving effects of EFFs affect the cost–benefit to farmers? This section discusses the ATT of the cost–benefit of EFFs using the PSM model; the estimated results are shown in

Table 10. ATT of cost–benefit (NNM).

		ATT	SE	T-Value
Fertilisers application amounts		−4.824 ***	1.619	−2.980
Costs	Fertilisers costs	149.606 ***	18.534	8.072
	Pesticides costs	−19.271 **	5.433	3.547
	Hired labour costs	−13.410 **	6.189	2.167
	MDTFT costs	45.971 **	20.256	2.270
	Total costs	133.853 *	76.559	1.748
Yield		32.652 **	15.992	2.042
Revenue		255.677 **	118.153	2.164
Net revenue		106.485 *	56.472	1.886

Note: Table 10 shows the ATT of the average cost–benefit of peanut per mu. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

.

EFFs have good economic benefits. Firstly, adopting EFFs resulted in a reduction in fertiliser application of 4.824 kg/mu, a reduction of 20.1%. This confirmed that EFFs prevent non-point source pollution and promoted green agriculture. However, the higher price of EFFs increased the cost of fertiliser application by 149.606 CNY/mu. At the same time, EFFs prompted farmers to adopt MDTFT, which increased the cost of MDTFT by 45.971 CNY/mu. In addition, EFFs reduced hired labour and pesticide costs by 13.41 CNY/mu and 19.271 CNY/mu, respectively. However, EFFs still increased the total cost by 133.853 CNY/mu. At the same time, EFFs increased the yield and revenue by 32.652 kg/mu and 255.677 CNY/mu, respectively. Ultimately, EFFs increased net revenue by 106.485 CNY/mu.

5. Discussion

EFFs are multifunction fertilisers, which can not only solve the problem of harmonising the ecological environment and agricultural production, but the spillover effect of the application effect can also change agricultural labour use. Existing studies have mainly focused on the field trial area and confirm the effect of fertiliser application reduction and ecological regulation function with EFFs [1,35]; however, a systematic analysis of the changes in agricultural labour inputs is lacking in the literature. This study bridges the gap in the existing literature and analyses the impact of EFFs on agricultural labour using the PSM method and ESR model to eliminate bias due to observable and unobserved factors. At the same time, the influence mechanisms affecting the labour-saving effect of EFFs were clarified from the perspective of the application reduction effect, the fertilisation technology effect, and the time effect. Analysing the substitution effects of EFFs on labour is of great practical importance to alleviate labour pressure in agriculture.

The result shows that the application of EFFs has a spillover effect on agricultural labour inputs. Usually, the adoption of new technologies affects agricultural labour resource allocation [22,24,25]. This study provides a more comprehensive analysis of the labour substitution of EFFs from the perspective of the number of labourers and labour days. Therefore, applying EFFs is essential to reduce agricultural labour costs and optimise household labour resource allocation. In general, the characteristics of new agricultural technologies can directly change agricultural labour use [36]. Consistent with previous research, EFFs reduce the FAA through technical characteristics, such as extending the fertiliser effect time and locking nutrients, thus reducing labour inputs. The difference is that EFFs have ecological regulation functions that enhance soil water and the fertility retention capacity as well as reduce the incidence of pests and diseases [4,10], thereby further enhancing the labour-saving effect. Thus, the technical characteristics and the ecological regulation effect of EFFs can significantly influence the labour-saving effect. The labour-saving effect will be further optimised by improving the ecological regulation of EFFs.

The ecological functions of EFFs have strong positive externalities and inter-period effects, which become essential factors affecting the labour-saving effects of EFFs. Compared with previous studies [25], the application of EFFs has systematic effects on the FAA, FAF, PAA, and IWC by improving the ecological environment of the soil, which ultimately reduces agricultural labour inputs. However, farmers have profit maximisation as the goal. Farmers tend to prioritise economic benefits over environmental benefits. If the positive externality is not compensated, it may limit the ecological function of EFFs, which in turn affects the labour-saving effect of EFFs. More importantly, the ecological regulating effect of EFFs on CLQ also requires a specific time cycle. If EFFs cannot be continuously applied for a long time, this will restrict enhancement of the CLQ and soil microenvironment. This situation will ultimately affect the substitution of EFFs for labour. Therefore, an ecological compensation policy should be formulated for the positive externalities of EFFs. At the same time, subsidies and incentives should be adopted to appropriately extend the application years of EFFs to give full play to the ecological regulation function and improve the labour-saving effect of EFFs.

Another notable finding is that EFFs promote the adoption of MDTFTs, further saving on agricultural labour inputs. Previous studies have examined the impact of a single new agricultural technology on the allocation of agricultural labour [23]. In contrast, this study found a relationship between EFFs and MDTFTs. After adopting EFFs, farmers are more likely to adopt MDTFTs to improve CLQ and fertiliser effectiveness of fertilisers. Moreover, MDTFTs not only directly replace labour but also reduce the amount and frequency of fertiliser applied by reducing fertiliser loss, ultimately increasing the labour-saving effect of EFFs. Thus, there is a self-reinforcing mechanism of the substitution effect of EFFs on labour through promoting the adoption of MDTFT. Hence, promoting MDTFTs according to local conditions will help reduce agricultural labour requirements.

This study provides a valuable reference for the impact of new agricultural technology adoption on labour input. However, limited by data availability, this study only utilised cross-sectional data, which have insufficient persuasive power for determining long-term effects. Although farmers were grouped based on application years to reflect the influence of time effects on the labour-saving effects of EFFs, additional analyses using long-term panel data is still needed in the future. In addition, EFFs affect both the total agricultural labour input and labour demand in different agriculture production segments. Therefore, additional investigations on the allocation of household labour resource are also necessary.

6. Conclusions

This study analysed the effects of EFFs on agricultural labour inputs using Chinese peanut farmers as an example, including the number of labourers and days of labour. The results indicated that EFFs have a significant labour-saving effect, reducing the number of labourers’ input and labour days in agriculture production by 0.127 persons/mu and 0.601 days/mu, reductions of 24.3% and 9.7%, respectively. This finding remained robust after several robustness tests. The mechanism indicated that the application reduction effect of EFFs reduces the FAA, FAF, PAA, and IWC, thereby saving agricultural labour inputs. The time effect reinforced this effect. More importantly, adopting EFFs helped farmers adopt MDTFT and promoted reduction in fertiliser use, indicating that causing EFFs have a self-deepening labour-saving mechanism. Scale heterogeneity existed in the labour substitution effect of EFFs, and the labour-saving effect after application by large-scale farmers was higher than that of small- and medium-sized farmers. Simultaneously, EFFs reduced fertiliser application but increased fertiliser application costs due to higher prices. In addition, EFFs increased MDTFT costs, reduced pesticide and hired labour costs, and ultimately increased total costs. Furthermore, EFFs increased the yields and returns, eventually increasing net farm returns.

The findings of this study have important policy implications. Firstly, the government should strengthen the publicity and scientific fertilisation technology training for farmers on EFFs and prioritise promoting EFFs among large-scale households to drive adoption by other farmers from point to point. Secondly, more attention should be paid to extending the application years of EFFs by subsidising the price of EFFs on a cooperative basis year by year. Thirdly, fertilisation patterns that match EFFs with MDTFTs should be promoted according to local conditions. For example, emphasis should be placed on the application of MDTFTs in eastern China, where the level of agricultural intensification is high and the terrain is flat, to enhance the labour-saving effect of EFFs.